[56129] in North American Network Operators' Group
why /8 announments are bad...
daemon@ATHENA.MIT.EDU (bmanning@karoshi.com)
Wed Feb 26 10:50:22 2003
From: bmanning@karoshi.com
To: cidr-report@potaroo.net (Geoff Huston)
Date: Wed, 26 Feb 2003 07:50:23 -0800 (PST)
Cc: vhsu@Covad.COM (Hsu Vicky), nanog@merit.edu, pfs@cisco.com
In-Reply-To: <5.1.0.14.2.20030227015633.01ed75f0@localhost> from "Geoff Huston" at Feb 27, 2003 01:57:07 AM
Errors-To: owner-nanog-outgoing@merit.edu
>
> In return, would Covad please consider performing some meaningful form of
> route
> aggregation or other measures to reduce the amount of noise that is being
> passed across the global routing tables that originates from Covad?
>
> http://www.cidr-report.org/cgi-bin/as-report?as=AS18566&view=4637
>
> suggests that Covad could withdraw some 483 BGP routing table entries,
> reducing the total number of entires originated by Covad from 490
> to an equivalent set of 8 aggregate routes.
perhaps this is not the time/place to raise the point,
but I'm coming to the conclusion that there is increasing
pushback to -NOT- announce space that is not in active use.
So-called "dark" space, i.e. the unused interstitial gaps
in delegated space that is the the product of sparse delegation
techniques, is perhaps more of a hazzard, esp. wrt. spam/traffic
generation than might have been considered in the past. think
forged source addresses...
if this is a rational line of argument, then two tactics present
themselves: 1) announce the individual, more specifics. this
has the effect of further bloating the routing table, incuring
the rath of the self-appointed routing table police (so watch out
Covad, don't do what Telstra did... :) 2) keep my number of
routing table entries consistant by "grooming" back my sparse
delegations into more homogenous groups, e.g. renumber folks in
the four /28s spread across the /19 into a single /26 - then
withdraw the /19 and announce the /26 in its place.
the number of routing table entries remains consistant and the
number of possible entries for forged source addresses is
dramatically reduced. Of course this will require a major rethink/
rewrite of most ISPs engineering practice/operating procedures,
as it will be much more common to see legitimate, long prefixs in
the routing system.
as usual, YMMV.
--bill
